Findit identified the culprit r557027 with confidence 70.0% in the config "chromium.win / Win7 Tests (dbg)(1)"
based on the flakiness trend:

https://findit-for-me.appspot.com/waterfall/flake?key=ag9zfmZpbmRpdC1mb3ItbWVyogELEhdNYXN0ZXJGbGFrZUFuYWx5c2lzUm9vdCJsY2hyb21pdW0ud2luL1dpbjcgVGVzdHMgKGRiZykoMSkvNjkyMTUvaW50ZXJhY3RpdmVfdWlfdGVzdHMvUzJWNVltOWhjbVJCWTJObGMzTlVaWE4wTGtKaFkydEdiM0ozWVhKa1MyVjVjdz09DAsSE01hc3RlckZsYWtlQW5hbHlzaXMYAQw


Automatically posted by the findit-for-me app (https://goo.gl/Ot9f7N).
Feedback is welcome! Please use component Tools>Test>FindIt>Flakiness

I've managed to get a local repro of this. I think it's just a classic race -- key presses are sent but it doesn't look like the test waits for the navigation to complete before checking the result. I'll see if I can make a quick fix.

Scratch that. It's not so simple. I've gotten far enough to know that I'm not the right person to carry this further. Maybe this has something to do with site isolation. With the logging I've added in https://chromium-review.googlesource.com/#/c/chromium/src/+/1073370 , I see different sets of mutations in the navigation controller for different failures.

In one case, I see this following the final forward press:

[12252:7136:0525/211607.846:ERROR:keyboard_access_browsertest.cc(484)] sending press
[12252:7136:0525/211607.849:ERROR:browser_commands.cc(435)] GoForward
[12252:7136:0525/211607.852:ERROR:navigation_controller_impl.cc(1883)] DiscardPendingEntry
[12252:7136:0525/211607.852:ERROR:navigation_controller_impl.cc(692)] GoToIndex
[12252:7136:0525/211607.853:ERROR:navigation_controller_impl.cc(1883)] DiscardPendingEntry
[12252:7136:0525/211607.908:ERROR:navigation_controller_impl.cc(1498)] RendererDidNavigateToExistingPage
[12252:7136:0525/211608.014:ERROR:keyboard_access_browsertest.cc(498)] after_forward:About Version
../../chrome/browser/ui/views/keyboard_access_browsertest.cc(503): error: Expected equality of these values:
  before_back
    Which is: L"Chrome URLs"
  after_forward
    Which is: L"About Version"

in another, I see this:

[14328:18548:0525/212802.525:ERROR:keyboard_access_browsertest.cc(444)] sending press
[14328:18548:0525/212802.532:ERROR:navigation_controller_impl.cc(682)] GoToIndex
[14328:18548:0525/212802.655:ERROR:keyboard_access_browsertest.cc(450)] after_back: About Version
[14328:18548:0525/212802.655:ERROR:keyboard_access_browsertest.cc(455)] sending press
[14328:18548:0525/212802.658:ERROR:browser_commands.cc(435)] GoForward
[14328:18548:0525/212802.662:ERROR:navigation_controller_impl.cc(1873)] DiscardPendingEntry
[14328:18548:0525/212802.662:ERROR:navigation_controller_impl.cc(682)] GoToIndex
[14328:18548:0525/212802.663:ERROR:navigation_controller_impl.cc(1873)] DiscardPendingEntry
[14328:18548:0525/212802.697:ERROR:navigation_controller_impl.cc(520)] SetPendingEntry
[14328:18548:0525/212802.735:ERROR:keyboard_access_browsertest.cc(461)] after_forward:chrome://version
../../chrome/browser/ui/views/keyboard_access_browsertest.cc(463): error: Expected equality of these values:
  before_back
    Which is: L"Chrome URLs"
  after_forward
    Which is: L"chrome://version"

Tentatively assigning to jam@ on account of some blame lines. Please note the high flake rate: https://test-results.appspot.com/dashboards/flakiness_dashboard.html#testType=interactive_ui_tests&tests=KeyboardAccessTest.BackForwardKeys. Thanks.

I'm curious why SiteIsolation label is on the bug, it seems just to be regular test flakiness, no?

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/70d35a0257713040ca2c506f78790c8ecedbc083

commit 70d35a0257713040ca2c506f78790c8ecedbc083
Author: Greg Thompson <grt@chromium.org>
Date: Fri May 25 20:32:53 2018

Disable KeyboardAccessTest.BackForwardKeys on debug due to flakes.

BUG=846623
TBR=tapted@chromium.org

Change-Id: Iafb9b3b179e35f4847d87c887a58b143524d0ce0
Reviewed-on: https://chromium-review.googlesource.com/1073372
Reviewed-by: Greg Thompson <grt@chromium.org>
Commit-Queue: Greg Thompson <grt@chromium.org>
Cr-Commit-Position: refs/heads/master@{#561996}
[modify] https://crrev.com/70d35a0257713040ca2c506f78790c8ecedbc083/chrome/browser/ui/views/keyboard_access_browsertest.cc

Adding a navigation label instead of Site Isolation, since I agree with Nasko that there doesn't seem to be anything Site Isolation related here.  Does seem to be a race with navigation that the test could probably fix by waiting for the navigation to complete.

sorry I'm too busy to investigate flakiness unless there's high likelihood that I caused it